Ritualized Behavior in the Middle Stone Age: Evidence from Rhino Cave, Tsodilo Hills, Botswana SHEILA COULSON

Department of Archaeology, Institute of Archaeology, Conservation, and History, Blindernveien 11 (Post Box 1019) University of Oslo, 0315 Oslo, NORWAY; [email protected]

SIGRID STAURSET

Department of Archaeology, Institute of Archaeology, Conservation, and History, University of Oslo, 0315 Oslo, NORWAY; [email protected]

NICK WALKER

8 Andrew Joss St., Mossel Bay 6506, SOUTH AFRICA; [email protected]

ABSTRACT Rhino Cave, located at the World Heritage site of Tsodilo Hills, is one of the three main Middle Stone Age (MSA) sites in Botswana. Initial investigations during the mid-1990s left unanswered a number of key questions regarding the early use of the cave. This prompted the current investigations, which have unearthed a wealth of MSA artifacts from a lag deposit. Results of a selectively employed chaîne opératoire analysis have revealed a very special set of behavioral patterns. It will be argued that the best-fit interpretation of the results from this investigation lies within the realm of ritualized behavior. The assemblage is characterized by an unexpectedly large number of MSA points, which are for the most part produced in non-locally acquired raw materials. These points are colorful, carefully and often elaborately made, and, once complete, never left the cave. They were either deliberately burned to the point where they could no longer be used, abandoned, or intentionally smashed. These artifacts were found together with tabular grinding slabs and pieces of the locally available pigment, specularite. This assemblage was recovered directly beneath a massive, virtually free-standing rock face that has been carved with hundreds of cupules of varying sizes and shapes. A section of the carved rock face was recovered from well within the MSA deposits in association with handheld grinding stones.

INTRODUCTION t has been proposed that “collective ritual – with its formal characteristic of amplified, stereotypical, redundant display – might be expected to leave a loud archaeological signature” (Watts 2009: 62). According to a number of researchers (e.g., Knight 2009; Knight 2010; Power 1999, 2009; Power and Aiello 1997; Watts 1999, 2002, 2009) the archaeological record of ochre use, dating from the late Middle Pleistocene, provides such a signature. Symbolism, a crucial feature of ritual behavior, has been central to the debate on behavioral modernity. Finds of shell beads (e.g., Bouzouggar et al. 2007; d’Errico et al. 2005; d’Errico et al. 2008; Henshilwood et al. 2004; Vanhaeren et al. 2006) and geometric engravings on red ochre (e.g., Barton 2005; d’Errico 2008; d’Errico et al. 2001; Henshilwood et al. 2002; Henshilwood et al. 2009; Henshilwood et al. 2001; Hovers et al. 2003; Jacobs et al. 2006; Mackay and Welz 2008; Mayer et al. 2009; Wilkins 2010) have been suggested to indicate the presence of symbolic traditions in the African Late Pleistocene, or Middle Stone Age (MSA). However, one area of investigation that could potentially reveal patterns of ritual behavior has received minimal attention within African MSA studies. This is the application of the chaîne

I

PaleoAnthropology 2011: 18−61. doi:10.4207/PA.2011.ART42

opératoire. Results from this method identify sequences of conscious choices made by the original artifact makers. The combined results of these individual sequences can provide insight into behavior patterns which are indicative of broader culturally determined traits and norms. By applying this methodological approach to a site that contains unusual or atypical MSA features, it can identify other ‘loud archaeological signatures’. As noted by McBrearty (2003: 514) “one of the most difficult tasks for the prehistorian is to identify the emergence of novel behaviors.” The authors suggest that a set of novel behavioral patterns is a most apt description for the results that have emerged from a recent test excavation at Rhino Cave, located within the World Heritage site of Tsodilo Hills, Botswana. This site was initially excavated in the mid-1990s (Robbins et al. 2000a; Robbins et al. 1996) and is one of the three main MSA sites in the country. These initial investigations focused primarily on the cave sediments and their potential for producing paleoenvironmental data. Analyses of the rich MSA archaeological assemblage, uncovered during these excavations, was limited to morphological descriptions and a typological classification of the tools. Although these initial investigations established

© 2011 PaleoAnthropology Society. All rights reserved.

ISSN 1545-0031

Ritualized Behavior in the Middle Stone Age • 19

Rhino Cave as an important archaeological site, a number of key questions regarding various aspects of the early use of the cave remain unanswered. Briefly, these include the unusual location of this site, the composition of the MSA assemblage, and the possible association of these finds to the rock art in this cave. Since these questions were the starting point of the present research study they will be presented and addressed in detail in the course of this paper (see section: Recent test excavation (2004–2006)). As will be described in the following pages, the recent investigations have unearthed a wealth of MSA artifacts from a lag deposit. Results of a selectively employed chaîne opératoire analysis revealed a very special and, to our knowledge, as yet unique set of behavioral patterns. The assemblage is characterized by an unexpectedly large number of MSA points, which are for the most part produced in non-locally acquired raw materials. These points are colorful, carefully and often elaborately made, and, once complete, have been treated in a manner which bears no resemblance to their normal repertoire of assigned hunting or cutting functions. They were either deliberately burned to the point where they could no longer be used, abandoned, or intentionally smashed. These points were found together with well-formed grinding stones, pigments, and tabular grinding slabs which, it will be proposed, could have been used to reduce specularite. Finally, this assemblage was recovered directly beneath a massive, virtually free-standing outcrop that dominates the interior of this small hidden cave. This outcrop has been carved with hundreds of cupules of varying sizes and shapes. In the following pages the geographical setting will be presented and a brief summary will be given of the procedures and strategies of the previous and recent excavations. The excavated assemblage will be introduced in light of the results of the chaîne opératoire analysis, where refitting was used to target aspects of this collection in an attempt to investigate questions regarding post-depositional disturbance, knapping patterns such as tool and blank selection, and specific behavioral choices. As this site has yet to be scientifically dated, emphasis has been placed on the MSA points since these artifacts are unquestionably fossiles directeurs for this period. Additional archaeological features from the MSA levels will also be presented, as will the results of night-time experiments to observe the effect of flickering light on the rock panel. Comparisons will be made to the lithic assemblages from Zimbabwe, Zambia, and Namibia, with special emphasis on the other main MSA sites in Botswana, White Paintings Shelter and ≠Gi. It will be argued that the best-fit interpretation of the results from the current investigation lies within the realm of ritualized behavior. GEOGRAPHICAL SETTING

berger and Hooper 1991: 2322). Occasionally this sand sea is broken by the occurrence of widely separated and heavily weathered rock outcrops, or inselbergs (Jacobberger and Hooper 1991: 2322). The most impressive and famous of these formations in Botswana are the Tsodilo Hills (Figure 1). Jutting from a seemingly unbroken flat landscape, Tsodilo Hills are the only major hills for over 100 km in any direction. They were formed in the Late Proterozoic and are composed of micaceous quartzites and quartz-schists (Cooke 1980: 82; Jacobberger and Hooper 1991: 2322; Thomas et al. 2003a: 54; Wright 1978: 239). Except for a series of small ephemeral springs and seeps in the three main hills, there is no surface water at Tsodilo Hills today (Thomas et al. 2003a: 54). However, as has been demonstrated in numerous studies, this was not always the case (e.g., Cooke 1975, 1980; Grove 1969; Thomas et al. 2003b; Thomas and Leason 2005; Thomas et al. 2000; Thomas and Shaw 1991, 2002). The Hills, as named by the local Hambukushu and Ju/’hoansi San, are called Male, Female, Child, and Grandchild (alternatively North Hill or “The Hill that Wants to Live by Itself”) (Campbell et al. 2010: 23, 177). Male Hill is the most dominant, rising to a height 410 meters from its immediate surroundings, making it the highest peak in Botswana (Jacobberger and Hooper 1991: 2322). The Hills also contain over 4,500 rock paintings which are located primarily on isolated rock panels and relatively unsheltered overhangs, constituting one of the highest concentrations of rock art in the world (ICOMOS 2001: 115). In 2001 the Hills were declared a World Heritage Site (ICOMOS 2001), although they had been protected under the Bushman Relics Act since 1934 (Segadika 2006: 31). The archaeological record of the area gives a account of human activities and environmental changes over at least 100,000 years (Robbins et al. 2000b). In addition to the rock art, the Hills also contain archaeological remains including Late and Middle Stone Age deposits (Robbins 1990; Robbins et al. 2000a; Robbins et al. 2000b; Robbins et al. 1996), Early Iron Age villages (Denbow 1986, 1990; Denbow and Wilmsen 1986) and numerous prehistoric specularite mines (Kiehn et al. 2007; Murphy et al. 2010; Robbins et al. 1998). Tsodilo may technically be a series of hills but it is regarded by near and distant communities as a mountain in spiritual terms, the home of the Gods. Even the names of the Hills articulate a personalized living idiom. The Hambukushu and Ju/’hoansi, who live or previously lived at the Hills, all have myths placing their origins here; several sites being the location of special rituals, including rainmaking and praying to the ancestors (see summary in Keitumetse et al. 2007: 108–109). Even today, many church groups make pilgrimages to the hills to pray and drink from its healing waters.

TSODILO HILLS The landscape of Botswana is covered by the degraded sand dunes, relict sand sheets, pans (ephemerally flooded depressions), and fossil river valleys of the Kalahari (Jacob-

RHINO CAVE Despite the fact that Tsodilo Hills and its surroundings are some of the most intensively investigated areas in Botswana, Rhino Cave escaped the notice of researchers un-

20 • PaleoAnthropology 2011

til the mid 1990s (Campbell et al. 1994b: 38). It was during the course of an audit of the rock art that Xhao Xontae, the headman of the local Ju/’hoansi San, revealed the existence of this cave to archaeologists (Xontae Xhao, son of Xhao Xontae and present headman of the local San, personal communication to contributor SC, 2007). It is, however, easy to understand how the site evaded detection, as it is perched high on the northernmost ridge of Female Hill and can only be approached by scrambling over, or squeezing between, massive boulders (Figure 1). Gaining entry to the cave is only slightly less arduous. On the western side of the ridge there is a raised, narrow, crawl space that ends with a considerable drop into the site. Alternatively, the wider eastern entrance offers two options: a two-meter jump or a slide down a steep boulder face, followed by a scramble over a rock-strewn opening near the present day floor (Figure 2). It is worth noting that during the MSA, the floor level would have been in excess of a meter lower and the entrance area further obstructed by large boulders (see Robbins et al. 2000a: Figure 4). Before the build-up of windblown sands on the eastern side of the Hills (Jacobberger and Hooper 1991: 2326; Lancaster 1981: 329–330) the climb to the top of the ridge would have presumably been even more difficult. The interior of Rhino Cave is formed by a narrow fissure in the quartzite host-rock that has created an opening which is just under 11m long, and varies from 1.25m to 5m wide, with a resultant floor area of about 22m² (see Figure 2). The high ceiling and walls of the cave extend beyond the boulders that dominate the eastern opening, effectively blocking any direct sunlight. The floor covering is flat, powdery and devoid of virtually any vegetation. Today, during heavy rains, water flows along the chamber behind the southern wall, washes across the entrance and eventually exits near the base of the sloping boulder in the north eastern part of the cave (Lopang Tatlhago, National Museum of Botswana employee, personal communication to contributor SC, 2005). Analyses of the deposits indicate standing water also would have collected in this part of the cave during the Early Holocene (Robbins et al. 2000a: 28; Robbins et al. 2000b: 1110). Clearly this is not an ideal habitation site, particularly as the Hills are replete with numerous well-protected rock shelters that are easily accessible. Yet Rhino Cave contains a number of obvious and distinctive features which attest to its use over a considerable period of time. Both main walls are decorated—the north wall contains a group of rock paintings, while virtually the entire quartzite outcrop, that dominates the south wall, has been ground to form an impressive series of grooves and depressions (see Figure 2). Furthermore, initial excavations at the site revealed Early Iron Age material and signs of sporadic use during the Late Stone Age (LSA), but quite surprisingly the weight of the evidence for use of the cave lies with the substantial body of lithic debitage attributable to the MSA (Robbins et al. 2000a; Robbins et al. 1996). The paintings on the north wall, which have been described in detail elsewhere (Robbins et al. 1996: 32–34), are

dominated by an animal that has been interpreted to be a white rhinoceros, which gives the cave its name. Alternatively, this figure could represent an elephant (Robbins et al. 2010: 60). The dominant feature of the south wall is a massive quartzite outcrop that is slightly under 7m long, 2m high and approximately 1m thick. As can be seen in Figure 2, it appears to be virtually free standing, as only the lower back section rests on the stone beneath it. Access to a chamber behind the outcrop is gained by crawling through the opening created by the upraised section closest to the main entrance of the cave. From this hidden vantage point it is possible to see into the cave through the 20–30cm gap that runs the full length of the top of the outcrop. By maneuvering onto the ledge that runs parallel to this gap (see Figure 2) a small-bodied person can work their way up to a narrow opening that leads outside the cave, although today this is choked by rocks. However, it is the face of this massive outcrop that is the immediate focus of attention upon entering the cave, as it has been ground with over three hundred grooves and depressions, hereafter referred to as cupules. These are confined to this single, vertical face and are concentrated on the lower 1.4m of the 2m high panel. The forms and condition of these cupules vary—some are long and thin, others oval to teardrop-shaped (also see Robbins et al. 1996: 34). They vary between 1–4cm in depth and while some are now heavily weathered, others appear to be relatively fresh. These cupules tend to overlap and truncate each other and many have been obliterated from repeated grinding. The profile of the panel suggests that several centimetres of rock have been removed by this action. These observations suggest that these cupules have been ground into the face of this outcrop over a long period of time. The carved face of this outcrop is illuminated during the mid-winter months, when for a few hours in the late afternoon a narrow arc of sunlight enters through a small opening in the ceiling of the cave and flickers directly across the carvings. EXCAVATIONS PREVIOUS INVESTIGATIONS (1995–1996) The excavations conducted in 1995 (Robbins et al. 1996) and 1996 (Robbins et al. 2000a) produced a wealth of materials predominantly attributable to the MSA. The publications of these investigations briefly describe the archaeological remains, with equal attention given to the paleoenvironmental data and the cave sediments. As questions arising from these investigations prompted the most recent test excavation, and because these excavation reports were published in journals that are not easily accessible, a brief summary will be given of the most important results. As can be seen in Figure 2, in 1995 and 1996 a trench was cut across the width of the cave, with two initial 1m excavation squares (Pits 1 and 2) extending from directly beneath the paintings on the north wall, followed by an extension in 1996 of two additional squares (Pits 3 and 4), terminating close to the gap under the carved panel. Very

Ritualized Behavior in the Middle Stone Age • 21

Figure 1. Map of Tsodilo Hills indicating the local archaeological sites referred to in the text. The hills are located approximately 40km west of the Okavango Panhandle (drawn by Sheila Coulson). little material was recovered from the pit nearest the paintings, a large boulder dominated most of the second pit and the large rock outcrop directly beneath the carved panel was encountered just 40cm into Pit 4, the southern limit of the 1996 extension. Therefore, the published accounts concentrate on the findings from the two central units (Pits 2 and 3) (Robbins et al. 2000a: 19; Robbins et al. 1996: 25), which were excavated to a depth of 140cm, although auger probes indicated an additional 160cm of as yet unexcavated deposit (Robbins et al. 1996: 25). The stratigraphy is composed of wedge-shaped units of deposit that broaden to the northwest and towards the interior of the cave (see stratigraphic cross section, Robbins et al. 2000a: Figure 5). The upper matrix is composed of medium to fine aeolian sand which is thought to have entered the site by way of gravity, water, or wind (Robbins et al. 2000a: 25, 29). These levels contained sparse indications of use of the cave during the most recent past and consist

of pottery, some lithic artifacts, fauna, and nut shell fragments. In Pit 2 these levels also contained denser concentrations of lithic materials from the LSA extending to a depth of 55cm. These loose sandy levels overlie a coarse deposit of sands and gravels created by the breakdown of the cave walls during colder, wetter periods (Robbins et al. 2000a: 27; Robbins et al. 1996: 28). These coarser deposits, which contain the materials attributed to the MSA, slope markedly. It is this slope that accounts for the MSA finds being concentrated below 90cm in Pit 2, while occurring at considerably shallower depths in Pit 3 (45cm) and Pit 4 (20cm). A morphological investigation of the MSA assemblage from the 1995–1996 excavations determined that there were 95 unifacial and bifacial MSA points, many of which exhibited indications of corner-struck (otherwise known as déjeté) manufacture (Robbins et al. 2000a: 17; Robbins et al. 1996: 32). Also recovered from these deposits were an array of over 250 retouched artifacts including scrapers, awls,

22 • PaleoAnthropology 2011

Figure 2. Rhino Cave, sketch map of the main features of the cave (photographs are taken from the cave floor and are not to scale). The paintings are immediately in front of the white mineral wash line, visible in far right, in the photo of the carved panel. The view of the entrance is completed by matching the slope on the large boulder to the continuation photo on the right (figure compiled by Sheila Coulson).

Ritualized Behavior in the Middle Stone Age • 23

denticulates, and burins, although almost half of this total is composed of miscellaneous retouched pieces and broken tools (Robbins et al. 2000a: 23; Robbins et al. 1996: 45). Hammerstones or grindstones also were recovered (Robbins et al. 2000a: 22; Robbins et al. 1996: 34). Of further note was the recovery of specularite, which, as defined by Kearey (1996: 292), is a platy, metallic variety of haematite. This was found in the form of metallic crystals and vein specularite, which was often rich enough in iron to be attracted by a magnet (Robbins et al. 2000a: 22). Approximately 12,000 pieces of lithic debitage were recovered from the MSA levels. As a dense concentration was found between the boulders in Pits 2 and 4 it was concluded that extensive “flint knapping” was conducted at this spot during the MSA (Robbins et al. 2000a: 29). The boulder, which fills most of Pit 2, was proposed to be a knapping seat, as next to it was found “some hammerstones and tools that were not finished…small retouching flakes and artifacts such as some MSA points with very sharp edges and tips that do not appear to have been used” (Robbins et al. 1996: 25). Elsewhere it is remarked that the points were recovered in various stages of manufacture, with some appearing “to have been abandoned as a result of failures at thinning or retouching, while others were finished and are in pristine condition” (Robbins et al. 1996: 32). Although an exact figure was not given for 1996, the initial excavations reported 56% of the raw materials used were locally available quartz and quartzite (Robbins et al. 1996: 39), with the remainder comprised of non-local materials including “several varieties of chert, jasper, chalcedony and silcrete” (Robbins et al. 2000a: 21). The presence of such “large amounts of exotic materials” is suggested as an indication that they were either acquired directly from their source or through exchange networks between the people at Tsodilo and those who resided near the various sources (Robbins and Murphy 1998: 61). The possible origins for these raw materials have been placed far afield, for example near the Aha Hills (>125km), along the Boteti River (>200km) and in the vicinity of Kudiakam Pan (>400km) (Murphy 1999: 372; Robbins 1987a, 1989; Robbins et al. 2000b: 1105; Weedman 1993: 68), although, as observed by Murphy (1999: 230), closer outcrops could have been exposed in the past. As noted by Robbins et al. (2000a) the deposits in Rhino Cave were heavily sloping and the upper levels were loose and powdery. However, the only specific mention of a control on the stratigraphic integrity for this site is the refitting of two pieces of a jasper flake, which were both found in the 55–60cm (LSA) level of Pit 2 (Robbins et al. 1996: 28). These factors should be taken into account when considering the reported “uncertain” and “problematic” radiocarbon and thermoluminescence (TL) dates for this site, specifically those from the MSA levels (Robbins et al. 2000a: 19, 29). A small sample of charcoal, taken from Pit 3 at 65–70cm, was used to obtain the first date of 14,500±50 BP for the MSA level. The second date is from a TL sample, which produced a date of 18,175±2871 BP, and was taken from the upper 30cm of the thin heterogeneous deposits of Pit 4. As the excavators admit, this is not only next to the

cave wall, but also within centimetres of the rock outcrop beneath it. Furthermore, this date was determined without field gamma spectrometer readings, meaning that the water content of the deposits and the annual dose at the sampling site was not measured (Robbins et al. 2000a: 19). The dating of the MSA assemblage from this site then had to revert to a typological comparison to the nearby welldated MSA sites of White Paintings Shelter (66,400±6,500 and 94,300±9,400 BP) (Robbins et al. 2000b: 1092) and the open-air pan site of ≠Gi (77,000±11,000 BP) (Brooks et al. 1990: 62) which is approximately 120km southwest of the Hills (Robbins and Murphy 1998: 60). For example Laurel Phillipson (2007: 20), who recently re-examined a selection of the MSA points from the 1996 excavations, states that she concurs “with the excavators that the Middle Stone Age lithics from these adjacent sites appear so similar… that the series of age determinations from the White Paintings Shelter can also be applied to the material from Rhino Cave.” Therefore, although the exact dating of Rhino Cave remains unresolved, this MSA assemblage can be considered to be generally comparable to the those from White Paintings Shelter and ≠Gi (Robbins and Murphy 1998: 60). The original excavators drew few conclusions as to how Rhino Cave was used. Due to the fairly small size of the cave it was determined that it would have only accommodated a small group of people (Robbins and Murphy 1998: 60) and on the basis of the amount of lithic debitage recovered they surmise that during the MSA the cave was used “over a lengthy period” of time (Robbins et al. 2000a: 29). The proximity of the cluster of LSA artifacts in Pit 2 to the painted wall is found to be “interesting” (Robbins et al. 2000a: 30), while the only definitive statement on the function of the cave is that the site may have been used for brief visits to conduct “rituals in connection with the rock art” during the Early Iron Age on the basis of pottery, nutshells, bone fragments, and a scattering of lithic finds, dated to that period (Robbins et al. 2000a: 30; Robbins et al. 1996: 35). However, the southern wall of the cave, with its hundreds of carved grooves and depressions, remained “enigmatic in relation to age and function” (Robbins et al. 2000a: 30). [Note: also see Brook et al. (2011) for minimum dates of two cupules.] RECENT TEST EXCAVATION (2004–2006) As Rhino Cave is only the third main MSA site in Botswana, it is not surprising that the initial investigations produced more questions than answers. The impetus for a new test in this cave was, therefore, a combination of tantalizing observations and unanswered questions that arose from these initial excavation results. One of the present contributors (NW) felt that testing closer to the carved wall of the cave would offer the greatest potential for determining how and when this panel was carved (Walker 2010). For the other contributors (SC and SS) their attention was drawn to the published statements pertaining to the extremely rich and unusual MSA lithic assemblage—for example, as locally available quartz was used in tool production, then why were large quantities of non-local raw materials be-

24 • PaleoAnthropology 2011

ing brought to this cave; were these materials used for restricted or specific purposes such as the production of certain types of artifacts; and, why were so many complete, unused, ‘pristine’ points reportedly manufactured and then left in this unlikely knapping site? However, the overriding unanswered question was still—how was this cave used during the MSA? In an attempt to further explore these deposits and to answer some of the outstanding questions which arose as a result of the earlier excavations, a 1m² test pit was positioned directly beneath the central portion of the carved wall and well away from the water run-off area (see Figure 2). A preliminary test was initiated here in 2003 (by contributor NW), but excavations began in earnest in 2004 (Coulson 2004)—the work reported in this paper is based on the materials from 2004–2006. The square was excavated in 50x50cm units and, as the upper deposit is composed of loose sands, the finds were not piece plotted. Initially the quadrants were excavated by bucket units (each bucket being 1/100th of a cubic meter), however, this proved to be difficult to control and the excavation of the quadrants was changed to digging in 5cm spits that were augmented with depth readings. It was soon evident that the only stable wall in the excavation was the rock outcrop directly beneath the carved panel. The other three profiles required the support of wooden shoring and account for why this test took so long to complete. In 2006, the west wall of the pit was extended by an additional two half quadrants (25x50cm each) to accommodate the extendable plywood shoring and to allow a rigid folding ladder to be opened in the pit, thereby offering an additional safety measure for the excavator (Coulson 2006). In addition, as these half quadrants were dug in 5cm spits, they served as a control for the finds retrieved from the bucket units during the 2004 investigations. Unfortunately, at various points during the excavations there were small wall collapses of these deposits. In each case the resultant material was treated separately and marked accordingly. Two members of the crew were assigned to watch for any indication of material filtering down behind the shoring: this was not only to ensure the integrity of the recovered materials but primarily to assure the safety of the person working in the bottom of the pit. This test pit was excavated to a depth of 185cm before the limit of the support shoring was reached, although archaeological material was still being recovered. As would be anticipated in such a small cave the upper section of deposit from the test pit corresponds with the loose sandy matrix reported from the 1995–1996 excavations (see Robbins et al. 2000a: Fig 5, Units A and B). However, in this part of the cave this forms a 65–90 cm thick layer of homogeneous sand—it is this loose powdery matrix that made the walls of the test pit so unstable. This deposit lies directly over an unmistakable densely packed lag deposit of coarse sands and gravels, exfoliated sections of the cave walls, and an astonishing amount of knapping debitage. This level is encountered between 80–100 cm below the surface and was still continuing with no visible change in composition

when digging stopped at 185cm (see Robbins et al. 2000a: Fig 5, Unit D). This deposit slopes towards the entrance of the cave, however, the slope is not as pronounced as was reported for the earlier excavations. The archaeological finds from the recent test excavation can be divided into three distinct groupings (as determined by contributor SS, Staurset 2008). The uppermost two are composed of material that can be assigned to the Early Iron Age or Late Stone Age (these will be described and discussed in a future publication). Below these layers, between 80–100 cm, is the aforementioned abrupt change in the deposit from sands to coarse sands and gravels. This lag deposit marks the appearance of finds attributable to the MSA. As can be seen in Table 1, these include unifacial and bifacial points, large scrapers, denticulates, becs, burin spalls, a variety of core types including Levallois and discoidal, and a large amount of knapping debitage (ca 10,000). In addition, this deposit contained a variety of handheld grinding stones, many of which are well-made, numerous lumps of specularite and small fragments of ochre. Unfortunately, faunal remains were not preserved. With a few exceptions, artifacts attributable to the MSA are confined to this deposit and despite the coarse matrix are in excellent condition. There was no obvious change in the assemblage, which was still continuing when excavations reached the limit of the support shoring at 185cm. Use of the locally available milky and glassy quartz and quartzites is evident throughout all of the layers, although there is an increase in the amount of good quality quartz from about 125cm. However, raw materials that are not found in the Hills, or the immediate vicinity (hereafter referred to as non-locally available or non-locally acquired raw materials), such as silcrete and a variety of silcretized materials, especially chalcedonies, were recovered in large numbers predominantly from the MSA deposits. It should be noted that the closest source of silcrete to Tsodilo Hills is approximately 50km to the west in the Xaudum Valley (Nash and Hopkinson 2004: 1543) and the nearest source of chalcedony is located in excess of 100km away (Monutsiwa Gabadirwe, Geologist with National Museum of Botswana, personal communication to contributor SC, 2008). Obviously, there is a need for reliable dates for the finds from Rhino Cave, especially with regard to the finds from the MSA deposits. This is recognized as a priority but has been deferred until the next stage of investigations at this site. Our prolonged excavation of a single pit, positioned adjacent to the cave wall, did not comply with the requirements for TL sampling (e.g., Liana and Roberts 2006; Richter 2007; Richter and Krbetschek 2006; Richter et al. 2007). However, the volume of debitage recovered indicates that there must have been a considerable amount of human activity over a prolonged period of time. ANALYSIS OF THE MSA ASSEMBLAGE FROM THE RECENT TEST EXCAVATION METHODOLOGY—THE CHAÎNE OPÉRATOIRE To accommodate the wealth of material retrieved from the

Ritualized Behavior in the Middle Stone Age • 25

TABLE 1. ARTIFACT COUNTS FOR MSA LEVEL OF RECENT TEST EXCAVATION 2004–2006.* Artifact types

Total 292

MSA points: Unifacial Partially bifacial Bifacial Other point types and point fragments: Levallois points Point fragments Misc. retouched Scrapers Denticulates Becs Burin spalls Grinding stones - handheld Cores: Levallois Discoidal Kombewa Amorphous Bipolar Single and double platform

53 2 33 5 5 55 23 5 4 5 28

21 14 1 23 5 10

*Note: the majority of the miscellaneous retouched pieces are probably unrefitted point fragments. They have been listed separately to maintain the integrity of the categories ‘points’ and ‘point fragments’.

excavation, a methodology was required that would not only provide a general overview of the entire collection but also enable detailed analyses to be undertaken on select groups of artifacts. The chaîne opératoire approach was chosen as it has proven to be highly successful on other sites containing assemblages where a specific set of questions were posed (e.g., Bergman et al. 1990; Bodu et al. 1987; Cahen and Keeley 1980; Cahen et al. 1979; Coulson 1986; Edmonds 1990; Skar and Coulson 1986). Although the chaîne opératoire is commonly used as a methodological approach for analyzing collections from European Stone Age sites, elements of this method have only recently been implemented in Southern Africa (e.g., Lombard 2005, 2006c; Lombard et al. 2004; Van Peer and Wurz 2006; Villa et al. 2005; Wurz 1999, 2000, 2005; Wurz et al. 2005). Its application to the MSA collections from Rhino Cave marks the first occasion of its use in Botswana. The growing popularity of the chaîne opératoire is partially due to its ability to surpass typological description by identifying the sequence of conscious choices made by the original artifact makers. The combined results of these individual sequences provide insight into behavior pat-

terns indicative of broader culturally determined traits and norms. The application of this methodological approach has resulted in the production of firm evidence for a variety of aspects of the MSA that in the past would have been beyond our reach. These include hunting and tool production procedures (Lombard 2004, 2005, 2006a, b, c, 2007; Lombard et al. 2004; Lombard and Wadley 2007; Soriano et al. 2009; Wadley 2005; Wadley et al. 2004), site use and management (Wadley 2006b; Wadley and Jacobs 2004; Wadley and Whitelaw 2006), and style (Wurz 1999, 2000, 2005). However, it is also possible to use the chaîne opératoire to provide evidence of a more fundamental nature. Even before the most recent testing began at Rhino Cave there were unanswered questions from the previous investigations, such as—to what degree had post-depositional movement affected the deposits (Robbins et al. 2000a: 19; Robbins et al. 1996: 25) and, in light of the exceptional quantities of lithic materials recovered from this cave, would it even be possible to separate the various individual sequences of artifact production (Robbins et al. 2000a: 19; Robbins et al. 1996: 29)? These questions needed to be answered before individual behavioral patterns, such as the proposed on-

26 • PaleoAnthropology 2011

To view this article with high resolution photographs, please download from the homepage for PaleoAnthropology for 2011 http://www.paleoanthro.org/journal/contents_dynamic.asp

Figure 3. Refitted partially bifacial MSA point from Rhino Cave with three flakes and flake fragments from the final stages of production. To the right are two associated flakes that could not be refitted but came from the same manufacturing sequence. Depths in centimeters below surface (BS) are listed. All artifacts are crenated or crazed from burning – the flake on the right is most affected (photo by Sheila Coulson). site manufacture of tools, could be examined. To address these fundamental concerns, as well as the research questions regarding behavior patterns, the method of refitting was applied. Refitting is a very simple method. The goal of a refitter is exactly the inverse of the prehistoric artifact maker, or knapper, “in that the intention is to reconstruct the original nodule from which flakes were struck” (Cahen 1987: 1). Whereas practically any site will benefit from refitting analysis, what will be achieved is totally dependent on the approach, in particular the questions being posed and the groups of materials selected. In this study (for contributors SS and SC) the initial focus of investigation at Rhino Cave was on the non-local lithic raw material types and specifically the MSA points produced from them. In preparation for analysis, the entire collection from the recent test pit was washed and sorted according to raw material type and the artifacts selected for further analysis were labelled. This included all complete, or virtually complete, retouched MSA points, as well as all debitage produced in the same raw material. These selected groups were placed on separate refitting trays, thereby allowing the groups to be easily compared and cross-checked. This is an essential step, for, as has been noted previously (Villa et al. 2005: 405), a refitting, or technological study, can not be conducted from within the confines of a finds bag. As can be seen for the example

shown in Figure 3 these materials were sufficiently distinct that it was a relatively simple process to refit the debitage and to determine whether or not the remaining flakes were likely to refit. Attempts to refit a group were stopped once it was determined that continued efforts would probably not produce any additional information. As a further control of the layer integrity, the corresponding finds from the LSA deposit were included in this analysis. No refits were possible between artifacts found in the LSA and MSA layers. EXTENT OF POST-DEPOSITIONAL MOVEMENT Before the research questions posed at the onset of this investigation regarding MSA behavioral patterns and the use of the cave could be addressed, it was first necessary to examine the question of post-depositional disturbance on this site. As previously noted, the loose sands of the upper levels would have easily allowed movement within the deposit. Even in the densely packed lower level we cannot discount bioturbation or other disturbance factors associated with lag deposits. As has been previously noted (e.g., Dibble et al. 1997; Richardson 1992; Villa 1982), even on sites with supposedly distinct stratigraphic layering this should not be taken as a guarantee that artifacts have remained fixed in their original locations. Fortunately, at

Ritualized Behavior in the Middle Stone Age • 27

Rhino Cave the technology and typology of the LSA and MSA materials are so distinct that most artifacts could be readily identified. This also meant that it was possible to determine, even during the course of excavation, that a few diagnostic MSA artifacts appear to have percolated up through the looser sediments of the LSA layer and must therefore be considered out of context. In an attempt to begin to determine the extent of disturbance in the MSA deposit from this site, artifacts were selected that would offer the best opportunity for refitting. Fortunately, this proved to be a surprisingly straightforward procedure as, for example, it was quickly determined that diagnostic artifacts, such as the MSA points, were made from a wide variety of distinctly colored materials, with easily distinguishable associated knapping debitage (for example, see Figure 3). This selection process also was aided by the overall excellent condition of the lithic assemblage. Contrary to reports by the previous excavators (Robbins et al. 1996: 32), there is virtually no indication of discoloration or patination on the lithic assemblage from this site. Reports by Phillipson (2007: 20) of the apparent “heavily worn, blunted and/or broken” condition of a selection of MSA points from the 1996 excavations has been responded to elsewhere (Coulson 2008). It was possible (for contributors SS and SC) to refit a wide variety of knapping groups at Rhino Cave. The majority form clusters of two or three flakes that were recovered either from within the same excavation unit or were separated horizontally or vertically by less than 10cm (the full refitting study will be presented in detail in a future publication). An example that offered evidence of the extent of movement in this deposit is shown in Figure 3. This partially bifacial MSA point has three refitted flakes or flake fragments which, when replotted, indicate 24cm of vertical displacement. Also illustrated are two flakes which were determined to have come from the same knapping sequence. This attribution was made on the basis of distinct technological features and an immediately recognizable raw material. These flakes are considered to be ‘associated’ to this refitted group and increase the level of vertical displacement to 40cm. It bears repeating that, as the excavation was conducted in 5cm spits, this is the maximum extent of separation possible—if piece plotting had been feasible on this site, the distance between these artifacts would most likely have been reduced. Finally, in addition to the two flakes shown in Figure 3, this associated group also contains 11 tiny chips. These chips were recovered from within the same 40cm of deposit as determined for the refitted group and the two associated flakes; therefore, refitting them would not have contributed additional information regarding the limits of disturbance. This range of disturbance from bioturbation is not unusual or excessive (Cahen and Moeyersons 1977; Hofman 1986). At the nearby rockshelter of White Paintings Shelter, disturbance was determined to range between 0–30cm, which the excavators felt was “modest” (Robbins et al. 2000b: 1091). While at the open-air MSA site of ≠Gi, Kuman (1989: 267) confirmed that the only way to determine the

exact level of vertical movement was through the replotting of sets of conjoined pieces. Unfortunately, at this site refitting was not possible because of the restricted sample size at her disposal. Therefore, the level of disturbance at ≠Gi was judged on the basis of artifact condition and size sorting, revealing that there had been considerable movement of the artifacts by trampling, which was exacerbated by slow soil build-up (Kuman 1989: 279). On the basis of the refitting analysis on groups of artifacts from the recent test excavation at Rhino Cave there is evidence to support a degree of movement within the coarse sands and gravels containing the MSA assemblage. Most refits indicate movement of between 0–10cm, with the maximum range being 40cm. This pattern was confirmed by the range of depth measurements from groups of obvious raw material types that have yet to be refitted but clearly represent a single tool and its associated waste flakes. These results combined with the plotting of diagnostic tool types indicate that some of the MSA material had percolated a few centimeters up into the sands, but there is no indication of LSA artifacts working their way down into the coarse sands and gravels. ANALYSIS OF THE MSA POINTS As has been previously stated the research interest for two of the contributors (SC and SS) focused on the extremely rich and unusual MSA lithic assemblage from this site. Initially this was in response to unanswered questions that arose from the findings from the 1995–1996 excavations. Results from the most recent test excavation only served to reinforce the relevance of investigating the role played by the broad range of non-local raw materials brought to this site and its connection to the overriding question of how this site was used. To address these questions the MSA points from the recent test excavation were selected for an in-depth chaîne opératoire analysis. The method of refitting was used in an attempt to determine manufacturing characteristics and to reveal production differences which could be attributed to the use of a wide variety of different raw material types. The analysis focused on the points because they were produced in both locally available and non-locally acquired raw materials and, unlike other tool types, such as scrapers or burins, their production is confined to the MSA ( 2). As there are, as yet, no dates for these deposits at this site, analysis of this particular artifact group would ensure our results would indicate behavioral patterns attributable to the MSA (McBrearty and Brooks 2000: 497–498, 500). For the purposes of this analysis all MSA points from within the 1m³ of deposit from the recent test excavation were investigated as a single unit. Future investigations may detect differences over time throughout the deposit but with an established range of 40cm of possible displacement there was no justification for attempting to separate these artifacts according to depth measurements. Finally, examples of MSA points from the 1995–1996 excavations have been included when it was deemed useful to illustrate various distinctive features. These are labelled as being

28 • PaleoAnthropology 2011

TABLE 2. ARTIFACT COUNTS FOR MSA LEVEL RECENT TEST EXCAVATION 2004–2006 (indicating locally available raw materials, non-locally available materials, and unburnt and burnt artifacts).* Artifact types MSA points: Local quartz + quartzite Non-local materials Other point types and point fragments: Levallois points Local quartz + quartzite Non-local materials Point fragments Local quartz + quartzite Non-local materials Misc. retouched Local quartz + quartzite Non-local materials Scrapers: Local quartz + quartzite Non-local materials Denticulates Local quartz + quartzite Non-local materials Becs Local quartz + quartzite Non-local materials Burin spalls Local quartz + quartzite Non-local materials Grinding stones - handheld Local quartzite and ironstone Cores: Levallois Local quartz + quartzite Non-local materials Discoidal Local quartz + quartzite Non-local materials Kombewa Local quartz + quartzite Non-local materials Amorphous Local quartz + quartzite Non-local materials Bipolar Local quartz + quartzite Non-local materials Single and double platform Local quartz + quartzite Non-local materials

Unburnt 224

Burnt 68

Total 292

26 36

0 26

26 62

2 3

0 0

2 3

0 0

0 5

0 5

9 24

0 22

9 46

8 14

0 1

8 15

2 3

0 0

2 3

1 3

0 0

1 3

1 4

0 0

1 4

28

0

28

2 13

0 6

2 19

6 7

1 0

7 7

0 1

0 0

0 1

10 8

1 4

11 12

5 0

0 0

5 0

4 4

0 2

4 6

*Note: the majority of the miscellaneous retouched pieces are most probably unrefitted point fragments. They have been listed separately to maintain the integrity of the categories ‘points’ and ‘point fragments.’

Ritualized Behavior in the Middle Stone Age • 29

To view this article with high resolution photographs, please download from the homepage for PaleoAnthropology for 2011 http://www.paleoanthro.org/journal/contents_dynamic.asp

Figure 4. Selection of Middle Stone Age points from Rhino Cave. All are from the recent test excavation with the exception of three from the 1995 excavation—these are from RC2 95-100 (top row, fourth from left and bottom row, center two) (photo by Sheila Coulson). from Rhino Cave Pit 2 or Pit 3 (RC2 or RC3). The MSA points from the recent test excavation can be subdivided as follows: 53 unifacial, 2 partially bifacial, and 33 bifacial points, plus 5 Levallois or discoidal pointed flakes, which by definition have no retouched edges. Even though this last category has been intentionally produced from prepared cores, and has proven to function more than adequately as spearheads (Boëda et al. 1999; Lombard 2005), they have nevertheless been removed from the following analysis as it could be argued they were not irrefutably meant to serve as points. This leaves a total of 88 retouched MSA points for this analysis. The predominant raw material type used in their manufacture is chalcedony (42), with silcrete (18) and an as yet unidentified shiny, white, silcretized organic material (2) accounting for the final examples of the non-locally acquired raw materials. The remainder are produced in locally available milky and glassy quartz (24) and quartzite (2). Thus 30% of the MSA points from this test pit were made from locally available materials while 70% were produced in materials that had to be brought to the cave from outside the Hills. As can be seen in Figure 4, these MSA points vary from short, thick and foliate, to triangular or oval-shaped, and display a wide range of styles. The length (determined from the tool’s base to tip) ranges from 24–66mm with an average length of 35mm. A common feature in their manufacture is that cortex is retained at the base, with a natural corticated concavity in the nodule being intentionally positioned along the central axis of the tool. An equally regular occurrence is that blanks were chosen from déjeté flakes

indicating their removal from a discoidally prepared core. There is little indication of tools being reworked—the majority are unifacial, and on the basis of the angle of the retouch, there appears to have only been one row of retouch applied to shape the point (e.g., Towner and Warburton 1990). This observation was confirmed by the results of the refitting analysis. A surprising feature, which immediately became evident during the process of sorting the raw materials in preparation for the refitting study (as determined by contributor SS), was that virtually every point appeared to be produced on a flake taken from a different core. As can be seen in any of the figures illustrating these points, including the artifacts from the earlier excavations (Robbins et al. 2000a: Figure 6; Robbins et al. 2010: Figure 3.19; Robbins and Murphy 1998: Plate 2a; Robbins et al. 1996: Figure 6) the range and variety of colors, particularly in the tools made from chalcedony, is quite exceptional. Even if allowance is made for variations in color within a single nodule or block of raw material, it could not account for this level of diversity. Furthermore, the cores retrieved from these MSA deposits do not match the size or the color variation of the points. Large flakes (as is seen in Figure 5), that were presumably brought to the site as blanks for tool production, also follow this pattern—they do not match the color variations for the points or the cores found on-site. The colors of these potential blanks are not a result of patination or staining—this range of vibrant colors reflects the natural shades of the raw materials. With regard to the silcretes this is harder to determine, as this material has a limited range

30 • PaleoAnthropology 2011

To view this article with high resolution photographs, please download from the homepage for PaleoAnthropology for 2011 http://www.paleoanthro.org/journal/contents_dynamic.asp

Figure 5. Selection of colorful flakes of non-local raw materials brought to Rhino Cave. Right of separation line: from the 1995–1996 excavations. Top row, second from right is from RC2 90-95, the remainder are from RC3 ranging between 60–85cm. Left of separation line: burnt flake of unusual breccia from the recent test excavation—the large mottled red and white flake adjacent to it is in the same material but unburnt (photo by Sheila Coulson). of colors, but by including features such as grain size, the debitage in this material also seems to follow the same pattern. Therefore, it appears the non-locally acquired raw materials brought to this cave represent a range of individual choices, or ‘one-offs,’ where bright coloring was a decisive factor in the selection process. The refitted knapping debitage for the MSA points made from non-locally acquired raw materials confirms that these tools were completed in the cave, as has been illustrated in the example in Figure 3. Once this pattern was determined, it was readily apparent that the vast majority of the remaining, as yet unrefitted, points and their associated waste materials followed this exact same pattern. This was also found to be the case for the points made from locally available milky and glassy quartz (Figure 6). Obviously refitting these materials (as determined by contributor SS) was considerably more difficult. However, it was possible to not only refit points from this material to their manufacturing debitage, but also, in a limited number of cases, to conjoin these points to a series of flake blanks that had been transformed into scrapers. Again, this indicates that the points and, in this case, associated other tool types, were not only manufactured in the cave but also that the tools and debitage had remained in the area where they were originally manufactured. Experimental replication of the production sequence for these point types has not yet been undertaken so it is

not possible to determine potential manufacturing pitfalls (for production failures in the MSA see Villa et al. 2009). However, producing the thinner, more fragile tip would be likely to be a source of failure. This is not the case at Rhino Cave, where the tips of the 88 MSA points recovered are complete or virtually complete. Furthermore, there are no indications of impact fracturing from use on these points. One of the contributors (SC) has experience with refitting this type of macro-fracture breakage from other research projects and is familiar with the groundbreaking experiments of Bergman and Newcomer (1983), Fischer et al. (1984) and Fischer (1989). The typical pattern of fluting or ‘burination’, especially at the tip, and bend and snap fractures common for this type of damage, were readily apparent on the artifacts examined from comparative collections such as White Paintings Shelter and ≠Gi, as reported by Donahue et al. (2004) and Brooks et al. (2006: 240). However, the few occurrences of minor damage observed on the MSA points for Rhino Cave can be attributed to minor edge damage from the coarse deposit, burning, or other sources of more recent damage. Further examination of the assemblage revealed that a large portion of the non-locally available material had been subject to burning—a chi-square test indicated this relationship is statistically significant (chi-square=42.663, degrees of freedom=1, p=